[FieldTrip] MEG sensors statistics

[Stephen Politzer-Ahles]

Hello Elana,

This depends a lot on what statistical test you did and how your data were
acquired and processed. e.g., to take the simplest example: if you simply
do t-tests at each sensor and don't correct for multiple comparisons, this
certainly is 'false power' (it's extremely anti-conservative: it will get a
lot of false positives). However, if you use cluster-based statistics, this
is less of a concern, since sensors showing related activity likely get
grouped together. The fieldtrip tutorial on cluster-based permutation
testing has more information.
Even with that, there still are potentially related issues, for instance,
if you recorded with axial gradiometers then a single dipole might show up
as two clusters of sensors, whereas if you recorded with planar
gradiometers (or transformed the data to the planar gradient) then a single
dipole should only show up as one cluster.

What you are describing in your message is also possible (it's similar to
calculating global field power, which people do do), but it throws away any
spatial information from the signal. Which analysis is most appropriate for
you really depends on the nature of the hypotheses you want to test.

Best,
Steve


---
Stephen Politzer-Ahles
University of Oxford
Language and Brain Lab
Faculty of Linguistics, Phonetics & Philology
http://users.ox.ac.uk/~cpgl0080/


>
> Message: 8
> Date: Mon, 29 Feb 2016 20:20:17 +0000
> From: "Harris, Elana" <Elana.Harris at cchmc.org>
> To: FieldTrip discussion list <fieldtrip at science.ru.nl>
> Cc: "Shahana, Nasrin \(Nasrin Shahana\)" <Nasrin.Shahana at cchmc.org>
> Subject: [FieldTrip] MEG sensors statistics
> Message-ID:
>         <
> B622DB18EE766C4BBAA34EA95310C34C5021AA68 at MCEXMB3.chmccorp.cchmc.org>
> Content-Type: text/plain; charset="utf-8"
>
> Dear All,
>
> I have the minimum, maximum and mean alpha power for each MEG sensor
> (found using MEGProcessor developed by Jing Xiang, MD, PhD) for 16
> subjects, 8 of whom have a psychiatric condition and 8 are healthy
> comparison subjects.
>
> We find statistical significance when comparing the set all of the MLF or
> MRO sensors, for instance, between the groups (healthy and pathological),
> however, I am wondering if this is false statistical power as all of the
> sensors likely represent once source generator and not truly, multiple,
> separate, non-independent recordings. Would it be best to summarize each
> subject?s data with one number to represent the maximum alpha power for the
> group of MLF sensors and then compare the two groups of subjects?
>
> I welcome your advice.
>
> Thank you.
>
> Elana
>
> Elana Harris, MD, PhD
> Assistant Professor
> Division of Child & Adolescent Psychiatry
> Cincinnati Children?s Hospital Medical Center
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <
> http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160229/598ee99d/attachment-0001.html
> >
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20160301/3da09bef/attachment-0001.html>